With more than 50% of all patients receiving radiation therapy (RT) for the management of their cancers, RT is an essential part of a successful cancer treatment. In recent years, the use of kV and MV imaging has greatly increased due to the adoption of image-guided radiation therapy techniques. Direct imaging of the tumor position while the MV treatment beam is ON has potential for improving tumor targeting, leading to better patient outcomes and reduced irradiation of healthy tissues. An alternative method employs real-time kV fluoroscopic tracking to perform 3D tumor-position tracking. However, this method can only be used for short periods of time as excess imaging dose can quickly exceed patient skin-dose tolerance levels. MV electronic portal imaging devices (EPID) have the distinct advantage that they can avoid this excess imaging dose by performing real-time tumor tracking using the actual MV treatment beam. However, MV EPIDs generally suffer from poor image quality. Typically, a thin (<1 mm) layer of scintillator is used to convert x-rays to visible light that is then detected by an array amorphous silicon (aSi) photodectors. These thin scintillators have very low detective quantum efficiency (DQE) and lead to images with poor contrast-to-noise ratios. We propose to overcome the EPID contrast obstacle by increasing the photon detection layer to 10-50mm using a transparent scintillator and then capturing its 4D light field using a specially designed optical camera system. The hypothesis is that by analyzing the 4D light field captured from a transparent scintillator, in-focus 2D planes along the MV beam direction can be effectively reconstructed. As contrast is linearly proportional to absorption efficiency, and absorption is linearly proportional to detector thickness, this technique has the potential to increase DQE by an order of magnitude over conventional techniques. The aims of the proposal are: Aim 1 ? Development of algorithms for simulating and processing MV based light fields. Aim 2 ? Experimental evaluation of a MV based light field camera. If successful, the approach will enable realtime and/or adaptive image-guided radiation therapy without addition of untargeted kilovoltage dose.